The evidence base for laboratory medicine: more work needed

The core purpose of laboratory medicine is to provide information that will help people either to maintain their health or in the event of illness to regain their health in so far as is possible with currently available treatments. Although laboratories generate vast amounts of information in the form of test results, the extent to which this confers benefit is often unclear.

Laboratory tests may be used in a variety of different ways: screening of asymptomatic subjects, diagnosis of disease, helping select the most appropriate form of treatment for a particular patient, monitoring disease progression or the effects of treatment and as a prognostic indicator. The potential benefits of testing include improved clinical outcomes, avoidance of unnecessary or inappropriate treatment along with psychological and other benefits related to timely diagnosis and intervention. Conversely, testing may be associated with adverse consequences. Most obviously, these might include detriment resulting from a false-positive or false-negative diagnostic test result but equally might include psychological harm to the subject tested and also social and economic detriment to wider society.

Despite the apparently inexorable rise in the workload of medical laboratories it is the case that for many tests there is little robust information available on their impact on patient outcomes and wellbeing. ^1,2 Assessment of laboratory tests has for too long focussed almost exclusively on reporting diagnostic accuracy in terms of the sensitivity and specificity with respect to the population studied. However, the mere fact of having available the result of a test does not in itself make any contribution to patient health or wellbeing. For a test result to confer benefit, someone, i.e. the clinician or the patient has to do something useful or effective with the result. For the clinician this may mean embarking on further tests, starting or adjusting treatment or for the patient engaging in some aspect of self-management behaviour, e.g. modification of lifestyle, improved concordance with prescribed medication. The value of a test cannot therefore be measured in isolation and must be considered in the context of the broader care pathway where the benefit to the patient may be some way downstream from the testing process. ² To quantify the benefit of a test, its specific impact must be extracted from a complex milieu which may be confounded by patient and disease heterogeneity, the role of other tests or investigations performed and the effectiveness of treatment.

The investigative and regulatory framework for assessing new tests may be compared to that in place for new therapeutic drugs (investigational medicinal products [IMPs]). IMPs are developed to treat a particular attribute or symptom of a disease process which can be defined and quantified. There is often a clear temporal relationship between the treatment and the desired outcome. Investigation of the effectiveness of an IMP readily lends itself to a prospective randomized controlled trial (RCT) design against placebo or comparator agent. By careful selection of inclusion and exclusion criteria it is possible to minimize the effect of confounding factors and to arrive at a clear answer on the efficacy of the new drug for the particular patient population studied (although the results may not have any validity beyond this carefully defined and selected cohort). IMP studies using hard clinical endpoints, e.g. cardiac event rates rather than surrogate event markers such as serum cholesterol concentration, may require large numbers of participants and depending on the disease in question, follow up for several years. IMP studies therefore tend to be very expensive but are generally funded and sponsored by the pharmaceutical company which holds the patent and which has a clear, vested, commercial interest in obtaining regulatory approval for the new drug.

An assessment of the clinical value of diagnostic tests is in many ways considerably more difficult due to the plethora of confounding factors and the gap between the testing process and clinical outcome. The most rigorous approach is the use of a ‘test-treatment’ RCT in which, in its simplest form, patients are randomized either to undergo the diagnostic test (test group) or not (non-test group). Test group patients then follow pre-determined clinical management pathways depending on whether the test result was positive or negative. ^1,2 Non-test group patients follow a conventional clinical management pathway. Clinical outcomes are compared between the test and non-test groups. It is clear that the impact on patient outcome may depend not only on the diagnostic accuracy of the test but also on the effectiveness of the treatment and the robustness of the overall clinical management pathway. As an extreme example, an accurate diagnostic test for a disease for which there are no effective treatments available may confer little overall benefit to the patient. Conversely, a poorly performing test combined with an effective treatment will also confer little additional benefit. The choice of the clinical outcomes to be evaluated also requires careful consideration and might include broader aspects of wellbeing such as the psychological impact of testing on subjects which may be particularly important in patients undertaking self-testing as part of chronic disease management. ³ It will be apparent that test-treatment RCTs are complex in design, may require large numbers of participants to attain adequate statistical power for the selected clinical endpoints and are therefore expensive to undertake. Alternative approaches include mathematical modelling which integrates results from diagnostic accuracy studies into clinical algorithms and computes outcomes. This approach has been most widely used in the design and assessment of screening programmes. ⁴

The evidence base for the role of laboratory testing for other applications such as monitoring of disease progression or the effect of treatment is even more scarce. Although monitoring is widely performed and the theoretical framework for evaluating the role of monitoring in different situations is well understood, there is often very little evidence base to support the monitoring strategies proposed in guidelines. ^5,6

Why have new laboratory tests not been evaluated with the same rigour as for new therapeutic drugs? There appear to be two main reasons. Firstly, there is not the same regulatory requirement for test manufacturers to demonstrate the clinical value of tests as is the case for the pharmaceutical industry. Secondly, the methodology required to evaluate tests is complex, requires integration of testing and treatment strategies and in consequence may be expensive. In the absence of any regulatory requirement there may be little incentive for test manufacturers to fund, wholly or even in part, the evaluations required. Such studies as are undertaken are generally supported by independent research funding which at least has the merit of minimizing concerns regarding transparency, openness and conflicts of interest that are often levelled at pharmaceutical industry sponsored research.

In the face of the challenge to find better evidence on the clinical value of laboratory tests there is an imperative for medical laboratory scientists and clinical teams to work collaboratively to formulate questions related to testing which are relevant to patients and to design and execute strategies to answer these questions. There is a need to generate both new evidence through further primary research and also to appraise critically existing evidence. Such considerations apply not only to new tests but to also established tests with new applications, HbA_1c (glycated haemoglobin) being an example of a monitoring test that was later adopted as a diagnostic test. ⁷ In their review article in this edition of the Annals, Price and Christenson discuss the importance of asking the right question as the first step in the evidence-based laboratory medicine cycle. ⁸ An appreciation of the knowledge gaps that are important to patients and their clinicians allows a question to be formulated using a PICO (Patient group of interest, Intervention to be assessed, Comparator intervention, Outcomes to be measured) framework. Everything in the subsequent analysis stems from a clear and precise articulation of the question to be answered: seeking the best evidence, appraising the evidence, applying it to clinical practice and analysing the impact. It is precisely this evidence base on the clinical value of testing that must inform best clinical practice.

In the UK, the National Institute for Health and Care Excellence (NICE) assesses available evidence on diagnostic tests through the Diagnostics Assessment Programme (DAP). ⁹ DAP evaluates ‘diagnostic technologies that have the potential to improve health outcomes but whose introduction is likely to be associated with an overall increase in cost to the NHS.’ This represents a valuable and important work programme with a focus very much on emerging tests and their cost-effectiveness compared to existing diagnostic approaches. Examples of guidance on laboratory tests recently published or in development include an assessment of two commercially available genetic tests for the diagnosis of familial hypercholesterolaemia and faecal calprotectin diagnostic tests for inflammatory diseases of the bowel. ^10,11 DAP will also evaluate screening tests in patients suspected of having disease (as opposed to tests used in population screening programmes which fall under the remit of the UK National screening Committee). ⁹ Companion diagnostic tests, i.e. to identify patients who respond best to new drugs are generally evaluated through the NICE Technology Appraisal programme alongside the drug to which it is related. ⁹

There has been a considerable improvement over the last decade in our understanding of the theoretical framework for evaluating the impact of laboratory testing. However, there is no room for complacency and it is essential that this knowledge is now applied if laboratory testing is to contribute to patient wellbeing in the most effective manner.